Employment vs investment

employment_vs_investment.py

from pylab import *
from numpy import *
import matplotlib.cbook as cbook
from datetime import datetime
from scipy.interpolate import interp1d

def read_file(filename):
    dates = []
    values = []
    for l in open(filename).readlines():
        datestr, valstr = l.strip().split()
        date = datetime.strptime(datestr, "%Y-%m-%d")
        dates.append(date)
        values.append(float(valstr))
    dates = date2num(dates)
    interpolated = interp1d(dates, array(values))
    d = date2num([ datetime(y,m,1) for y in range(1948,2009) for m in [1,4,7,10]])
    return (d, interpolated(d))

def year_as_num(year):
    return date2num(datetime(year,1,1))

t, employment = read_file("/tmp/EMRATIO.tsv")
t, investment = read_file("/tmp/GPDI.tsv")
t, gdp = read_file("/tmp/GDP.tsv")
investment_to_gdp = 100*investment/gdp

def color(time):
    if time < year_as_num(1980):
        return 'g'
    if time < year_as_num(1990):
        return 'r'
    return 'b'

scatter(employment, investment_to_gdp, c = [color(time) for time in t] )
ylabel("Investment to GDP (%)")
xlabel("Employment to Population (%)")

## a, b = polyfit(employment, investment_to_gdp,1)
## plot( employment, a*employment+b,'y', label="All years")

#Best fit line for 90's
idx = where(t >= year_as_num(1990))
a, b = polyfit(employment[idx], investment_to_gdp[idx], 1)
plot( employment[idx], a*employment[idx]+b, 'b', label="1990-2009") 

#Best fit for 80's
idx = where(logical_and(t >= year_as_num(1980), t <= year_as_num(1990)))
a, b = polyfit(employment[idx], investment_to_gdp[idx], 1)
plot( employment[idx], a*employment[idx]+b, 'r', label="1980-1990") 

#best fit for pre-80's
index = where(t <= year_as_num(1980))
a, b = polyfit(employment[index], investment_to_gdp[index], 1)
plot( sort(employment[index]), a*sort(employment[index])+b, 'g', label="1948-1980") 

legend()

show()